Sound source separating device, speech recognizing device, portable telephone, and sound source separating method, and program

Abstract

The sound source signal from a target sound source is separated from the mixed sound in which the sound source signals from sound sources are mixed without being influenced by the variations of the sensitivity of microphone elements are provided. A beam former section (3) of the sound source separating device (1) performs beam formation to attenuate the sound source signals arriving from directions symmetrical with respect to the vertical line of the line connecting two microphones (10, 11) by spectrum-analyzing the output signals from the microphones (10, 11) and multiplying the signals after the spectrum analysis by the weighting factors complex conjugated with the signals. Power calculating sections (40, 41) calculate power spectrum information. Target sound spectrum extracting sections (50, 51) extract spectrum information on target sound sources according to the difference between the power spectrum information from one beam former and that from the other.

Description

technical field [0001] The present invention relates to a sound source separating device for separating a sound source signal coming from a target sound source using a plurality of microphones, a plurality of audio signals emitted from a plurality of sound sources, or a signal mixed with a plurality of sound signals such as various environmental noises , an audio recognition device, a mobile phone, a sound source separation method, and a program. Background technique [0002] When you want to record a specific audio signal in various environments, since there are various noise sources in the surrounding environment, it is difficult to record only the target sound signal with a microphone, and some kind of noise reduction processing or sound source separation processing is required . [0003] As an example where these processes are particularly required, for example, a case in an automobile environment can be cited. In the automotive environment, due to the popularity of mo...

AI Technical Summary

Problems solved by technology

However, in GSC, there is a problem that when a target signal is transmitted from a direction deviated from the set target sound source direction, the target signal is canceled and degraded.

[0022] Therefore, in this adaptive filter operation, when both the target sound and the specific noise are audio, it is necessary to detect the state in which both are speaking, and the higher the frequency of both speaking, stop estimating the adaptive filter. The frequency of the filter increases, and as a result, there is a problem that the noise removal performance of the adaptive filter decreases

[0023] In addition, in the situation where large diffuse noise such as automobile running noise is generated, the estimation accuracy of the adaptive filter is reduced due to the influence of the diffuse noise. As a result, there is the following problem: The estimation accuracy of the speaker direction and the direction of arrival of the specific noise direction of the estimated filter parameters is reduced, and the overall performance of the system is reduced.

[0024] Furthermore, in the above-mentioned methods described in Patent Documents 6 and 7, when the noise component suppressing device is configured using a low-cost microphone for an automobile or the like, the initial reflection of the sound source in the vehicle is large, and the diffuse noise component is large. , The deviation of the microphone element used is large (about ±3db), the driver and the passenger speak at the same time, and there is a highly correlated target sound and specific noise at the same time, the adaptive filter unit does not perform the desired operation , cannot achieve the desired action as a whole

Images